Whispera Voice Commands (MLX)

This repo hosts a small MLX-compatible model fine-tuned to convert natural-language (voice-like) macOS commands into structured JSON:

{"category":"apps","operation":"open","app":"chrome"}

That JSON is intended to be mapped to real shell commands using the templates/patterns in macos_operations.json from the main project repo.

Project: https://github.com/sapoepsilon/whisperaModel

Usage (mlx_lm)

mlx_lm.generate --model sapoepsilon/whispera-voice-commands --prompt "open safari" --max-tokens 100 --temp 0.1

Notes

  • This is a fused (merged) model for convenience (base + LoRA adapters).
  • Outputs are JSON only; you should parse the first {...} block and then apply your own command templates.
Downloads last month
142
Safetensors
Model size
0.5B params
Tensor type
BF16
·
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support